2 Copyright (C) 2000-2005 Kern Sibbald
4 This program is free software; you can redistribute it and/or
5 modify it under the terms of the GNU General Public License as
6 published by the Free Software Foundation; either version 2 of
7 the License, or (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 General Public License for more details.
14 You should have received a copy of the GNU General Public
15 License along with this program; if not, write to the Free
16 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
19 This file is based on GNU TAR source code. Except for a few key
20 ideas, it has been rewritten for Bacula.
24 Thanks to the TAR programmers.
36 extern int32_t name_max; /* filename max length */
37 extern int32_t path_max; /* path name max length */
40 * Structure for keeping track of hard linked files, we
41 * keep an entry for each hardlinked file that we save,
42 * which is the first one found. For all the other files that
43 * are linked to this one, we save only the directory
44 * entry so we can link it.
48 dev_t dev; /* device */
49 ino_t ino; /* inode with device is unique */
51 uint32_t FileIndex; /* Bacula FileIndex of this file */
52 char name[1]; /* The name */
55 static void free_dir_ff_pkt(FF_PKT *dir_ff_pkt)
57 free(dir_ff_pkt->fname);
58 free(dir_ff_pkt->link);
59 free_pool_memory(dir_ff_pkt->sys_fname);
64 * Check to see if we allow the file system type of a file or directory.
65 * If we do not have a list of file system types, we accept anything.
67 static int accept_fstype(FF_PKT *ff, void *dummy) {
72 if (ff->fstypes.size()) {
74 if (!fstype(ff->fname, fs, sizeof(fs))) {
75 Dmsg1(50, "Cannot determine file system type for \"%s\"\n", ff->fname);
77 for (i = 0; i < ff->fstypes.size(); ++i) {
78 if (strcmp(fs, (char *)ff->fstypes.get(i)) == 0) {
79 Dmsg2(100, "Accepting fstype %s for \"%s\"\n", fs, ff->fname);
83 Dmsg3(200, "fstype %s for \"%s\" does not match %s\n", fs,
84 ff->fname, ff->fstypes.get(i));
92 * This function determines whether we can use getattrlist()
93 * It's odd, but we have to use the function to determine that...
94 * Also, the man pages talk about things as if they were implemented.
96 * On Mac OS X, this succesfully differentiates between HFS+ and UFS
97 * volumes, which makes me trust it is OK for others, too.
99 static bool volume_has_attrlist(const char *fname)
101 #ifdef HAVE_DARWIN_OS
103 struct volinfo_struct {
104 unsigned long length; /* Mandatory field */
105 vol_capabilities_attr_t info; /* Volume capabilities */
107 struct attrlist attrList;
109 memset(&attrList, 0, sizeof(attrList));
110 attrList.bitmapcount = ATTR_BIT_MAP_COUNT;
111 attrList.volattr = ATTR_VOL_INFO | ATTR_VOL_CAPABILITIES;
112 if (statfs(fname, &st) == 0) {
113 /* We need to check on the mount point */
114 if (getattrlist(st.f_mntonname, &attrList, &vol, sizeof(vol), FSOPT_NOFOLLOW) == 0
115 && (vol.info.capabilities[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_ATTRLIST)
116 && (vol.info.valid[VOL_CAPABILITIES_INTERFACES] & VOL_CAP_INT_ATTRLIST)) {
125 * Find a single file.
126 * handle_file is the callback for handling the file.
128 * parent_device is the device we are currently on
129 * top_level is 1 when not recursing or 0 when
130 * descending into a directory.
133 find_one_file(JCR *jcr, FF_PKT *ff_pkt,
134 int handle_file(FF_PKT *ff, void *hpkt, bool top_level),
135 void *pkt, char *fname, dev_t parent_device, bool top_level)
137 struct utimbuf restore_times;
141 ff_pkt->fname = ff_pkt->link = fname;
143 if (lstat(fname, &ff_pkt->statp) != 0) {
144 /* Cannot stat file */
145 ff_pkt->type = FT_NOSTAT;
146 ff_pkt->ff_errno = errno;
147 return handle_file(ff_pkt, pkt, top_level);
150 Dmsg1(300, "File ----: %s\n", fname);
152 /* Save current times of this directory in case we need to
153 * reset them because the user doesn't want them changed.
155 restore_times.actime = ff_pkt->statp.st_atime;
156 restore_times.modtime = ff_pkt->statp.st_mtime;
159 * We check for allowed fstypes at top_level and fstype change (below).
162 if (!accept_fstype(ff_pkt, NULL)) {
163 ff_pkt->type = FT_INVALIDFS;
164 if (ff_pkt->flags & FO_KEEPATIME) {
165 utime(fname, &restore_times);
167 Jmsg1(jcr, M_ERROR, 0, _("Top level directory \"%s\" has an unlisted fstype\n"), fname);
168 return 1; /* Just ignore this error - or the whole backup is cancelled */
170 ff_pkt->volhas_attrlist = volume_has_attrlist(fname);
174 * If this is an Incremental backup, see if file was modified
175 * since our last "save_time", presumably the last Full save
178 if (ff_pkt->incremental && !S_ISDIR(ff_pkt->statp.st_mode)) {
179 Dmsg1(300, "Non-directory incremental: %s\n", ff_pkt->fname);
180 /* Not a directory */
181 if (ff_pkt->statp.st_mtime < ff_pkt->save_time
182 && ((ff_pkt->flags & FO_MTIMEONLY) ||
183 ff_pkt->statp.st_ctime < ff_pkt->save_time)) {
184 /* Incremental option, file not changed */
185 ff_pkt->type = FT_NOCHG;
186 return handle_file(ff_pkt, pkt, top_level);
190 #ifdef HAVE_DARWIN_OS
191 if (ff_pkt->flags & FO_HFSPLUS && ff_pkt->volhas_attrlist
192 && S_ISREG(ff_pkt->statp.st_mode)) {
193 /* TODO: initialise attrList once elsewhere? */
194 struct attrlist attrList;
195 memset(&attrList, 0, sizeof(attrList));
196 attrList.bitmapcount = ATTR_BIT_MAP_COUNT;
197 attrList.commonattr = ATTR_CMN_FNDRINFO;
198 attrList.fileattr = ATTR_FILE_RSRCLENGTH;
199 if (getattrlist(fname, &attrList, &ff_pkt->hfsinfo,
200 sizeof(ff_pkt->hfsinfo), FSOPT_NOFOLLOW) != 0) {
201 ff_pkt->type = FT_NOSTAT;
202 ff_pkt->ff_errno = errno;
203 return handle_file(ff_pkt, pkt, top_level);
208 /* ***FIXME*** implement this */
210 /* See if we are trying to dump the archive. */
211 if (ar_dev && ff_pkt->statp.st_dev == ar_dev && ff_pkt->statp.st_ino == ar_ino) {
212 ff_pkt->type = FT_ISARCH;
213 return handle_file(ff_pkt, pkt, top_level);
218 * Handle hard linked files
220 * Maintain a list of hard linked files already backed up. This
221 * allows us to ensure that the data of each file gets backed
224 if (!(ff_pkt->flags & FO_NO_HARDLINK)
225 && ff_pkt->statp.st_nlink > 1
226 && (S_ISREG(ff_pkt->statp.st_mode)
227 || S_ISCHR(ff_pkt->statp.st_mode)
228 || S_ISBLK(ff_pkt->statp.st_mode)
229 || S_ISFIFO(ff_pkt->statp.st_mode)
230 || S_ISSOCK(ff_pkt->statp.st_mode))) {
234 /* Search link list of hard linked files */
235 for (lp = ff_pkt->linklist; lp; lp = lp->next)
236 if (lp->ino == (ino_t)ff_pkt->statp.st_ino &&
237 lp->dev == (dev_t)ff_pkt->statp.st_dev) {
238 /* If we have already backed up the hard linked file don't do it again */
239 if (strcmp(lp->name, fname) == 0) {
240 Jmsg1(jcr, M_WARNING, 0, _("Attempt to backup hard linked file %s twice ignored.\n"),
242 return 1; /* ignore */
244 ff_pkt->link = lp->name;
245 ff_pkt->type = FT_LNKSAVED; /* Handle link, file already saved */
246 ff_pkt->LinkFI = lp->FileIndex;
247 return handle_file(ff_pkt, pkt, top_level);
250 /* File not previously dumped. Chain it into our list. */
251 len = strlen(fname) + 1;
252 lp = (struct f_link *)bmalloc(sizeof(struct f_link) + len);
253 lp->ino = ff_pkt->statp.st_ino;
254 lp->dev = ff_pkt->statp.st_dev;
255 bstrncpy(lp->name, fname, len);
256 lp->next = ff_pkt->linklist;
257 ff_pkt->linklist = lp;
258 ff_pkt->linked = lp; /* mark saved link */
260 ff_pkt->linked = NULL;
263 /* This is not a link to a previously dumped file, so dump it. */
264 if (S_ISREG(ff_pkt->statp.st_mode)) {
267 sizeleft = ff_pkt->statp.st_size;
269 /* Don't bother opening empty, world readable files. Also do not open
270 files when archive is meant for /dev/null. */
271 if (ff_pkt->null_output_device || (sizeleft == 0
272 && MODE_RALL == (MODE_RALL & ff_pkt->statp.st_mode))) {
273 ff_pkt->type = FT_REGE;
275 ff_pkt->type = FT_REG;
277 rtn_stat = handle_file(ff_pkt, pkt, top_level);
278 if (ff_pkt->linked) {
279 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
284 } else if (S_ISLNK(ff_pkt->statp.st_mode)) { /* soft link */
286 char *buffer = (char *)alloca(path_max + name_max + 102);
288 size = readlink(fname, buffer, path_max + name_max + 101);
290 /* Could not follow link */
291 ff_pkt->type = FT_NOFOLLOW;
292 ff_pkt->ff_errno = errno;
293 rtn_stat = handle_file(ff_pkt, pkt, top_level);
294 if (ff_pkt->linked) {
295 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
300 ff_pkt->link = buffer; /* point to link */
301 ff_pkt->type = FT_LNK; /* got a real link */
302 rtn_stat = handle_file(ff_pkt, pkt, top_level);
303 if (ff_pkt->linked) {
304 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
308 } else if (S_ISDIR(ff_pkt->statp.st_mode)) {
310 struct dirent *entry, *result;
315 dev_t our_device = ff_pkt->statp.st_dev;
317 bool volhas_attrlist = ff_pkt->volhas_attrlist; /* Remember this if we recurse */
320 * If we are using Win32 (non-portable) backup API, don't check
321 * access as everything is more complicated, and
322 * in principle, we should be able to access everything.
324 if (!have_win32_api() || (ff_pkt->flags & FO_PORTABLE)) {
325 if (access(fname, R_OK) == -1 && geteuid() != 0) {
326 /* Could not access() directory */
327 ff_pkt->type = FT_NOACCESS;
328 ff_pkt->ff_errno = errno;
329 rtn_stat = handle_file(ff_pkt, pkt, top_level);
330 if (ff_pkt->linked) {
331 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
337 /* Build a canonical directory name with a trailing slash in link var */
339 link_len = len + 200;
340 link = (char *)bmalloc(link_len + 2);
341 bstrncpy(link, fname, link_len);
342 /* Strip all trailing slashes */
343 while (len >= 1 && link[len - 1] == '/')
345 link[len++] = '/'; /* add back one */
349 if (ff_pkt->incremental &&
350 (ff_pkt->statp.st_mtime < ff_pkt->save_time &&
351 ff_pkt->statp.st_ctime < ff_pkt->save_time)) {
352 /* Incremental option, directory entry not changed */
353 ff_pkt->type = FT_DIRNOCHG;
355 ff_pkt->type = FT_DIRBEGIN;
358 * Note, we return the directory to the calling program (handle_file)
359 * when we first see the directory (FT_DIRBEGIN.
360 * This allows the program to apply matches and make a
361 * choice whether or not to accept it. If it is accepted, we
362 * do not immediately save it, but do so only after everything
363 * in the directory is seen (i.e. the FT_DIREND).
365 rtn_stat = handle_file(ff_pkt, pkt, top_level);
366 if (rtn_stat < 1) { /* ignore or error status */
370 /* Done with DIRBEGIN, next call will be DIREND */
371 if (ff_pkt->type == FT_DIRBEGIN) {
372 ff_pkt->type = FT_DIREND;
376 * Create a temporary ff packet for this directory
377 * entry, and defer handling the directory until
378 * we have recursed into it. This saves the
379 * directory after all files have been processed, and
380 * during the restore, the directory permissions will
381 * be reset after all the files have been restored.
383 Dmsg1(300, "Create temp ff packet for dir: %s\n", ff_pkt->fname);
384 FF_PKT *dir_ff_pkt = (FF_PKT *)bmalloc(sizeof(FF_PKT));
385 memcpy(dir_ff_pkt, ff_pkt, sizeof(FF_PKT));
386 dir_ff_pkt->fname = bstrdup(ff_pkt->fname);
387 dir_ff_pkt->link = bstrdup(ff_pkt->link);
388 dir_ff_pkt->sys_fname = get_pool_memory(PM_FNAME);
389 dir_ff_pkt->included_files_list = NULL;
390 dir_ff_pkt->excluded_files_list = NULL;
391 dir_ff_pkt->excluded_paths_list = NULL;
392 dir_ff_pkt->linklist = NULL;
395 * Do not descend into subdirectories (recurse) if the
396 * user has turned it off for this directory.
398 * If we are crossing file systems, we are either not allowed
399 * to cross, or we may be restricted by a list of permitted
402 if (ff_pkt->flags & FO_NO_RECURSION) {
403 ff_pkt->type = FT_NORECURSE;
405 } else if (!top_level && parent_device != ff_pkt->statp.st_dev) {
406 if(!(ff_pkt->flags & FO_MULTIFS)) {
407 ff_pkt->type = FT_NOFSCHG;
409 } else if (!accept_fstype(ff_pkt, NULL)) {
410 ff_pkt->type = FT_INVALIDFS;
413 ff_pkt->volhas_attrlist = volume_has_attrlist(fname);
416 /* If not recursing, just backup dir and return */
418 rtn_stat = handle_file(ff_pkt, pkt, top_level);
419 if (ff_pkt->linked) {
420 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
423 free_dir_ff_pkt(dir_ff_pkt);
424 ff_pkt->link = ff_pkt->fname; /* reset "link" */
425 if (ff_pkt->flags & FO_KEEPATIME) {
426 utime(fname, &restore_times);
431 ff_pkt->link = ff_pkt->fname; /* reset "link" */
434 * Descend into or "recurse" into the directory to read
435 * all the files in it.
438 if ((directory = opendir(fname)) == NULL) {
439 ff_pkt->type = FT_NOOPEN;
440 ff_pkt->ff_errno = errno;
441 rtn_stat = handle_file(ff_pkt, pkt, top_level);
442 if (ff_pkt->linked) {
443 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
446 free_dir_ff_pkt(dir_ff_pkt);
451 * Process all files in this directory entry (recursing).
452 * This would possibly run faster if we chdir to the directory
453 * before traversing it.
456 entry = (struct dirent *)malloc(sizeof(struct dirent) + name_max + 100);
457 for ( ; !job_canceled(jcr); ) {
461 status = readdir_r(directory, entry, &result);
462 if (status != 0 || result == NULL) {
463 // Dmsg2(99, "readdir returned stat=%d result=0x%x\n",
464 // status, (long)result);
467 ASSERT(name_max+1 > (int)sizeof(struct dirent) + (int)NAMELEN(entry));
469 /* Skip `.', `..', and excluded file names. */
470 if (p[0] == '\0' || (p[0] == '.' && (p[1] == '\0' ||
471 (p[1] == '.' && p[2] == '\0')))) {
475 if ((int)NAMELEN(entry) + len >= link_len) {
476 link_len = len + NAMELEN(entry) + 1;
477 link = (char *)brealloc(link, link_len + 1);
480 for (i=0; i < (int)NAMELEN(entry); i++) {
484 if (!file_is_excluded(ff_pkt, link)) {
485 rtn_stat = find_one_file(jcr, ff_pkt, handle_file, pkt, link, our_device, false);
486 if (ff_pkt->linked) {
487 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
496 * Now that we have recursed through all the files in the
497 * directory, we "save" the directory so that after all
498 * the files are restored, this entry will serve to reset
499 * the directory modes and dates. Temp directory values
500 * were used without this record.
502 handle_file(dir_ff_pkt, pkt, top_level); /* handle directory entry */
503 if (ff_pkt->linked) {
504 ff_pkt->linked->FileIndex = dir_ff_pkt->FileIndex;
506 free_dir_ff_pkt(dir_ff_pkt);
508 if (ff_pkt->flags & FO_KEEPATIME) {
509 utime(fname, &restore_times);
511 ff_pkt->volhas_attrlist = volhas_attrlist; /* Restore value in case it changed. */
513 } /* end check for directory */
516 * If it is explicitly mentioned (i.e. top_level) and is
517 * a block device, we do a raw backup of it or if it is
518 * a fifo, we simply read it.
520 #ifdef HAVE_FREEBSD_OS
522 * On FreeBSD, all block devices are character devices, so
523 * to be able to read a raw disk, we need the check for
524 * a character device.
525 * crw-r----- 1 root operator - 116, 0x00040002 Jun 9 19:32 /dev/ad0s3
526 * crw-r----- 1 root operator - 116, 0x00040002 Jun 9 19:32 /dev/rad0s3
528 if (top_level && (S_ISBLK(ff_pkt->statp.st_mode) || S_ISCHR(ff_pkt->statp.st_mode))) {
530 if (top_level && S_ISBLK(ff_pkt->statp.st_mode)) {
532 ff_pkt->type = FT_RAW; /* raw partition */
533 } else if (top_level && S_ISFIFO(ff_pkt->statp.st_mode) &&
534 ff_pkt->flags & FO_READFIFO) {
535 ff_pkt->type = FT_FIFO;
537 /* The only remaining types are special (character, ...) files */
538 ff_pkt->type = FT_SPEC;
540 rtn_stat = handle_file(ff_pkt, pkt, top_level);
541 if (ff_pkt->linked) {
542 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
547 int term_find_one(FF_PKT *ff)
549 struct f_link *lp, *lc;
552 /* Free up list of hard linked files */
553 for (lp = ff->linklist; lp;) {